Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1256524 > unrolled thread

Re: [PATCH v3 7/7] PCI: hv: New paravirtual PCI front-end for Hyper-V VMs

Started byJiang Liu <jiang.liu@linux.intel.com>
First post2015-10-27 08:20 +0100
Last post2015-10-28 03:10 +0100
Articles 3 — 2 participants

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.


Contents

  Re: [PATCH v3 7/7] PCI: hv: New paravirtual PCI front-end for Hyper-V  VMs Jiang Liu <jiang.liu@linux.intel.com> - 2015-10-27 08:20 +0100
    RE: [PATCH v3 7/7] PCI: hv: New paravirtual PCI front-end for Hyper-V  VMs Jake Oshins <jakeo@microsoft.com> - 2015-10-27 22:00 +0100
      Re: [PATCH v3 7/7] PCI: hv: New paravirtual PCI front-end for Hyper-V  VMs Jiang Liu <jiang.liu@linux.intel.com> - 2015-10-28 03:10 +0100

#1256524 — Re: [PATCH v3 7/7] PCI: hv: New paravirtual PCI front-end for Hyper-V VMs

FromJiang Liu <jiang.liu@linux.intel.com>
Date2015-10-27 08:20 +0100
SubjectRe: [PATCH v3 7/7] PCI: hv: New paravirtual PCI front-end for Hyper-V VMs
Message-ID<qo6I2-2i5-21@gated-at.bofh.it>
On 2015/10/27 7:15, jakeo@microsoft.com wrote:
> From: Jake Oshins <jakeo@microsoft.com>
> 
> This patch introduces a new driver which exposes a root PCI bus whenever a PCI
> Express device is passed through to a guest VM under Hyper-V. The device can
> be single- or multi-function. The interrupts for the devices are managed by an
> IRQ domain, implemented within the driver.
> 
> Signed-off-by: Jake Oshins <jakeo@microsoft.com>
> ---
<snit>
> +
> +/**
> + * hv_pcie_init_irq_domain() - Initialize IRQ domain
> + * @hbus:	The root PCI bus
> + *
> + * Return: '0' on success and error value on failure
> + */
> +static int hv_pcie_init_irq_domain(struct hv_pcibus_device *hbus)
> +{
> +	hbus->msi_info.chip = &hv_msi_irq_chip;
> +	hbus->msi_info.chip_data = hbus;
> +	hbus->msi_info.ops = &hv_msi_ops;
> +	hbus->msi_info.flags = (MSI_FLAG_USE_DEF_DOM_OPS |
> +		MSI_FLAG_USE_DEF_CHIP_OPS | MSI_FLAG_MULTI_PCI_MSI |
> +		MSI_FLAG_PCI_MSIX);
When interrupt remapping is not supported, x86 vector allocator
can't support multiple MSI because it can't allocate continuous
vectors yet. So please confirm whether we could enable
MSI_FLAG_MULTI_PCI_MSI for HV.

> +	hbus->msi_info.handler = handle_edge_irq;
> +	hbus->msi_info.handler_name = "edge";
> +	hbus->msi_info.data = hbus;
How about using following pattern so we could avoid exporting
too many interfaces?

struct irq_domain *parent_domain = NULL;
hbus->msi_info.chip = &hv_msi_irq_chip;
hbus->msi_info.ops = &hv_msi_ops;
// Let arch code to fill in default ops for chip and domain
x86_setup_default_msi_irqdomian_info(&hbus->msi_info, &parent_domain);
// Override default ops if not applicable
hbus->irq_domain = pci_msi_create_irq_domain(hbus->fwnode,
					     &hbus->msi_info,
					     parent_domain);

> +	hbus->irq_domain = pci_msi_create_irq_domain(hbus->fwnode,
> +						     &hbus->msi_info,
> +						     x86_vector_domain);
> +	if (!hbus->irq_domain) {
> +		pr_err("Failed to build an MSI IRQ domain\n");
> +		return -ENODEV;
> +	}
> +
> +	return 0;
> +}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [next] | [standalone]


#1257192

FromJake Oshins <jakeo@microsoft.com>
Date2015-10-27 22:00 +0100
Message-ID<qojvB-1tZ-27@gated-at.bofh.it>
In reply to#1256524
> -----Original Message-----
> From: Jiang Liu [mailto:jiang.liu@linux.intel.com]
> Sent: Tuesday, October 27, 2015 12:11 AM
> To: Jake Oshins <jakeo@microsoft.com>; gregkh@linuxfoundation.org; KY
> Srinivasan <kys@microsoft.com>; linux-kernel@vger.kernel.org;
> devel@linuxdriverproject.org; olaf@aepfle.de; apw@canonical.com;
> vkuznets@redhat.com; tglx@linutronix.de; Haiyang Zhang
> <haiyangz@microsoft.com>; marc.zyngier@arm.com;
> bhelgaas@google.com; linux-pci@vger.kernel.org
> Subject: Re: [PATCH v3 7/7] PCI: hv: New paravirtual PCI front-end for Hyper-
> V VMs
> 
> On 2015/10/27 7:15, jakeo@microsoft.com wrote:
> > From: Jake Oshins <jakeo@microsoft.com>
> >
(snip)
> > +/**
> > + * hv_pcie_init_irq_domain() - Initialize IRQ domain
> > + * @hbus:	The root PCI bus
> > + *
> > + * Return: '0' on success and error value on failure
> > + */
> > +static int hv_pcie_init_irq_domain(struct hv_pcibus_device *hbus)
> > +{
> > +	hbus->msi_info.chip = &hv_msi_irq_chip;
> > +	hbus->msi_info.chip_data = hbus;
> > +	hbus->msi_info.ops = &hv_msi_ops;
> > +	hbus->msi_info.flags = (MSI_FLAG_USE_DEF_DOM_OPS |
> > +		MSI_FLAG_USE_DEF_CHIP_OPS |
> MSI_FLAG_MULTI_PCI_MSI |
> > +		MSI_FLAG_PCI_MSIX);
> When interrupt remapping is not supported, x86 vector allocator
> can't support multiple MSI because it can't allocate continuous
> vectors yet. So please confirm whether we could enable
> MSI_FLAG_MULTI_PCI_MSI for HV.
> 

We can actually handle the remapping in the hypervisor.  I'll add a comment to that effect.

> > +	hbus->msi_info.handler = handle_edge_irq;
> > +	hbus->msi_info.handler_name = "edge";
> > +	hbus->msi_info.data = hbus;
> How about using following pattern so we could avoid exporting
> too many interfaces?
> 
> struct irq_domain *parent_domain = NULL;
> hbus->msi_info.chip = &hv_msi_irq_chip;
> hbus->msi_info.ops = &hv_msi_ops;
> // Let arch code to fill in default ops for chip and domain
> x86_setup_default_msi_irqdomian_info(&hbus->msi_info,
> &parent_domain);
> // Override default ops if not applicable
> hbus->irq_domain = pci_msi_create_irq_domain(hbus->fwnode,
> 					     &hbus->msi_info,
> 					     parent_domain);
> 

I understand your point here, but I'm having trouble making it play out.  When I look at this, the only functions or structures which are supplied straight from exports (in my proposed patches) are irq_chip_ack_parent(), pci_msi_prepare() and x86_vector_domain.  The other exports either already exist for other reasons or they're needed within functions that I need to supply.  If you feel strongly that adding a new function to avoid exporting these is the right way to go, I'll do it, but I want to confirm that first.

Thanks,
Jake Oshins
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]


#1257578

FromJiang Liu <jiang.liu@linux.intel.com>
Date2015-10-28 03:10 +0100
Message-ID<qoolA-4PZ-9@gated-at.bofh.it>
In reply to#1257192
On 2015/10/28 4:38, Jake Oshins wrote:
>> -----Original Message-----
>> From: Jiang Liu [mailto:jiang.liu@linux.intel.com]
>> Sent: Tuesday, October 27, 2015 12:11 AM
>> To: Jake Oshins <jakeo@microsoft.com>; gregkh@linuxfoundation.org; KY
>> Srinivasan <kys@microsoft.com>; linux-kernel@vger.kernel.org;
>> devel@linuxdriverproject.org; olaf@aepfle.de; apw@canonical.com;
>> vkuznets@redhat.com; tglx@linutronix.de; Haiyang Zhang
>> <haiyangz@microsoft.com>; marc.zyngier@arm.com;
>> bhelgaas@google.com; linux-pci@vger.kernel.org
>> Subject: Re: [PATCH v3 7/7] PCI: hv: New paravirtual PCI front-end for Hyper-
>> V VMs
>>
>> On 2015/10/27 7:15, jakeo@microsoft.com wrote:
>>> From: Jake Oshins <jakeo@microsoft.com>
>>>
> (snip)
>>> +/**
>>> + * hv_pcie_init_irq_domain() - Initialize IRQ domain
>>> + * @hbus:	The root PCI bus
>>> + *
>>> + * Return: '0' on success and error value on failure
>>> + */
>>> +static int hv_pcie_init_irq_domain(struct hv_pcibus_device *hbus)
>>> +{
>>> +	hbus->msi_info.chip = &hv_msi_irq_chip;
>>> +	hbus->msi_info.chip_data = hbus;
>>> +	hbus->msi_info.ops = &hv_msi_ops;
>>> +	hbus->msi_info.flags = (MSI_FLAG_USE_DEF_DOM_OPS |
>>> +		MSI_FLAG_USE_DEF_CHIP_OPS |
>> MSI_FLAG_MULTI_PCI_MSI |
>>> +		MSI_FLAG_PCI_MSIX);
>> When interrupt remapping is not supported, x86 vector allocator
>> can't support multiple MSI because it can't allocate continuous
>> vectors yet. So please confirm whether we could enable
>> MSI_FLAG_MULTI_PCI_MSI for HV.
>>
> 
> We can actually handle the remapping in the hypervisor.  I'll add a comment to that effect.
> 
>>> +	hbus->msi_info.handler = handle_edge_irq;
>>> +	hbus->msi_info.handler_name = "edge";
>>> +	hbus->msi_info.data = hbus;
>> How about using following pattern so we could avoid exporting
>> too many interfaces?
>>
>> struct irq_domain *parent_domain = NULL;
>> hbus->msi_info.chip = &hv_msi_irq_chip;
>> hbus->msi_info.ops = &hv_msi_ops;
>> // Let arch code to fill in default ops for chip and domain
>> x86_setup_default_msi_irqdomian_info(&hbus->msi_info,
>> &parent_domain);
>> // Override default ops if not applicable
>> hbus->irq_domain = pci_msi_create_irq_domain(hbus->fwnode,
>> 					     &hbus->msi_info,
>> 					     parent_domain);
>>
> 
> I understand your point here, but I'm having trouble making it play out.  When I look at this, the only functions or structures which are supplied straight from exports (in my proposed patches) are irq_chip_ack_parent(), pci_msi_prepare() and x86_vector_domain.  The other exports either already exist for other reasons or they're needed within functions that I need to supply.  If you feel strongly that adding a new function to avoid exporting these is the right way to go, I'll do it, but I want to confirm that first.
No strong preference here, just feeling that pci_msi_prepare() and
x86_vector_domain are arch specific:)

And I was hoping that we could find some commonality between this
driver and the VMD driver posted at http://lkml.org/lkml/2015/10/27/607
so we could abstract the common part into core.

For the VMD driver, it has put the code to create irqdomain into core
as vmd_create_irq_domain().

Thanks,
Gerry

> 
> Thanks,
> Jake Oshins
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web